Ventilation rate prediction for naturally ventilated greenhouses using CFD-driven machine learning model

Despite the decrease in Korean agricultural land, the area of greenhouses is increasing due to an increase in demand for stable crop production.
In order to maintain high productivity in greenhouses, internal air environments such as temperature and humidity and carbon dioxide must be maintained under appropriate growth conditions.
In particular, properly controlled natural ventilation is an efficient and economical means of maintaining an appropriate internal air environment because additional energy is not consumed.
Computational fluid dynamics (CFD) techniques have actively been used for quantitative analyses and predictions of natural ventilation.
However, limited simulation cases can be usually performed because a considerable computation time is required for CFD modeling.
Conversely, techniques based on machine learning (ML) models have relatively shorter computation times but the construction of training data sets requires significant effort.
Therefore, the purpose of this study is to develop ML models to predict the natural ventilation rate by zone at the height of the crop group for a greenhouse and used CFD computed results to create a training data set.
The CFD simulation was performed considering external wind direction and wind speed as well as vent opening conditions.
The ventilation rate computed by the tracer gas decay (TGD) method was computed for 27 zones of the greenhouse.
Multiple regression, random forest, support vector regression, and deep neural network models were developed to predict the ventilation rate.
The training data set of wind direction and speed were supplemented, and accuracy analysis was conducted to improve the accuracy of each model; the number of data was supplemented by applying the bootstrapping technique to complement the limitations of a finite number of CFD cases.
The validity of developing ventilation rate prediction ML models using CFD was assessed by evaluating the accuracy of the developed ML models.

International Symposium on New Technologies for Sustainable Greenhouse Systems: GreenSys2023

H.H. Jeong, J.H. Cho, Y.B. Choi, S.M. Kang, D.I. Kim, Y.W. Cho, I.B. Lee

computational fluid dynamics, greenhouse, machine learning, naturally ventilated greenhouse, ventilation rate

https://doi.org/10.17660/ActaHortic.2025.1426.59